1. Human Health Risk Assessment: EPA’s Current Challenges and the Future Stan Barone Jr., PhD., National Center for Environmental Assessment Office of Research and Development United States Environmental Protection Agency Presentation for the National Capital Area Chapter - Society of Toxicology “Challenges and Opportunities in Putting High-Throughput Chemical Risk Characterization Into Real-World Practice” April 19, 2011 Washington, DC

2. Office of Research and Development National Center for Environmental Assessment 1 Human Health Risk Assessment Now and in the future, risk assessment remains fundamental to U.S. EPA’s approach to analysis of potential risk from exposure to environmental contaminants Essential for U.S. EPA regulatory decision-making Evolving in the face of new understandings about uncertainty, mode of action, metabolism, susceptibility, etc. Addressing emerging science and new science challenges 1

3. Table 4-31. Noncancer effects in animals repeatedly exposed to chemical x by the oral route Reference/species Exposure (mg/kg- day) NOAELLOAEL Effect (mg/kg-day) Burek et al., 1980 F344 rat, M&F 0, 0.05, 0.2, 1, 5, or 20 90 days in DW 0.2 1 5 1 5 20 Degenerative nerve changes Hindlimb foot splay Decreased body weight Atrophy of testes & skeletal muscle Johnson et al., 1986 F344 rat, M&F 0, 0.01, 0.1, 0.5, or 2.0 2 years in DW 0.5 2 0.5 2 ND 2 ND Degenerative nerve changes (L Hindlimb foot splay Decreased body weight Early mortality after 24 weeks Other nonneoplastic lesions Friedman et al., 1995 F344 rat, M&F 0, 0.1, 0.5, or 2.0 (M) 0, 1.0, or 3.0 (F) 2 years in DW 0.5(M) 1.0(F) 2.0(M) 3.0(F) 2.0(M) 3.0(F) ND degenerative nerve changes (L Decreased body weight (8 – 9%) Early mortality after 60 weeks Other nonneoplastic lesions Large number of animals Low throughput Expensive Time consuming Pathology endpoints Dose response extrapolations over a wide range Application of uncertainty factors Little focus on mode of action and biology Few epidemiology studies Current Approach 52

4. Office of Research and Development National Center for Environmental Assessment 3 Basic Principles of Risk Assessment at EPA The starting point for risk assessment is a critical analysis of available scientific information. Quantitative estimates of risk are, to the extent possible, – Biologically-motivated, – Data-driven. When there is insufficient data, default methods are used that – Protect public health, – Ensure scientific validity (i.e., scientifically plausible and extensively peer reviewed), and – Create an orderly, transparent and predictable process. Implementation of these principles involves extensive independent peer review.

5. Office of Research and Development National Center for Environmental Assessment 4 Human Health Risk Assessment Transforming to address emerging science and new science challenges There are tens of thousands of chemicals that are untested and lack assessment of potential for human toxicity. Current toxicology testing methods are too expensive, too slow, and can cope with too few chemicals. Toxicology approaches are evolving away from reliance on in vivo testing of laboratory animals Current approaches to risk analysis need to be significantly modified to deal with more chemicals; innovative approaches –Screening –Fingerprinting Risk assessment approaches must be developed that can use the new generation of data types and arrays; “omics” Thus, the environmental health community needs to develop next generation of risk assessment tools, approaches, and practices… NexGen risk assessment – Toxicity pathways – Focused high-throughput assessments

6. Office of Research and Development National Center for Environmental Assessment 5 Human Health Assessment Issues Mechanistic Considerations in Human Health Risk Assessment Increased need to characterize: – A wider array of hazard traits – More chemicals (no data on most chemicals in commerce) Human carcinogens increasingly emphasis on: – Multiple toxicity pathways, mechanisms affected – These mechanisms could inform new predictive approaches  In vitro assays  Human biomarkers Dose-response curve: – In an individual: can take multiple forms depending on genetic background, target tissue, internal dose – In a population: variability in susceptibility in response are key determinants Source: Guyton et al. Improving prediction of chemical carcinogenicity by considering multiple mechanisms and applying toxicogenomic approaches. Mutat Res. 681(2-3):230-40, 2009.

7. What Can Be Learned from Mechanistic Data and Analyses? Identify mechanism-based sources of human variability/ susceptibility (e.g., background diseases and processes, genetic polymorphisms, age, co- exposures) Address mechanism-based likelihood of other outcomes Improve prediction of interactions across environmental and endogenous exposures Identify mechanistic drivers of response at low-doses An individual’s dose response Background Exposure: Endogenous & Xenobiotic Heterogeneity in Background Exposure and Susceptibility Population dose response Environmental Chemical Dose Probability of Effect from Environmental Exposure Fraction of Population Responding to Environmental Chemical Environmental Chemical Stressor Adverse endpoint Biological Susceptibility: Health and Disease Status, Genetics, Age, Gender Source: National Academy of Sciences Report “Science and Decisions: Advancing Risk Assessment” Adapted from Figure 5-3a (December 2008) 21 6

8. Office of Research and Development National Center for Environmental Assessment 7 Increases appreciation of individual and population heterogeneity of disease mechanisms Improves prediction of interactions across environmental exposures Addresses mechanism-based likelihood of other outcomes Identifies mechanism-based sources of human variability/susceptibility (e.g., background diseases and processes, genetic polymorphisms, age, co- exposures) Uses Systems biology level tools and data Advances high throughput methodologies (microarray, proteomics) The use of mechanistic data will play a key role in the future of risk assessment to: –Aid in identification of sources of human variability/susceptibility (e.g., background diseases and processes, co-exposures, etc) and early stage disease biomarkers. –Address likelihood of other outcomes –Improve prediction of interactions across environmental and endogenous exposures –Indentify mechanistic drivers of response at low doses. Focus on Mechanisms of Human Disease

9. Office of Research and Development National Center for Environmental Assessment 8 Human Relevance/ Cost/Complexity Throughput/ Simplicity High-Throughput Screening Assays ( EPA’s National Center for Computational Toxicology, Office of Research and Development) 10s-100s/yr 10s-100s/day 1000s/day 10,000s- 100,000s/day LTSHTSMTSuHTS batch testing of chemicals for pharmacological/toxicological endpoints using automated liquid handling, detectors, and data acquisition Gene-expression

10. Office of Research and Development National Center for Environmental Assessment 9 Future of Toxicity Testing Bioinformatics/ Machine Learning in silico analysis Cancer ReproTox DevTox NeuroTox PulmonaryTox ImmunoTox HTS -omics in vitro testing $Thousands

11. Office of Research and Development National Center for Environmental Assessment 10 Toxicity Pathways Receptors / Enzymes / etc. Direct Molecular Interaction Pathway Regulation / Genomics Cellular Processes Tissue / Organ / Organism Tox Endpoint Chemical

12. ToxCast in vitro HTS assays Cell lines –HepG2 human hepatoblastoma –A549 human lung carcinoma –HEK 293 human embryonic kidney Primary cells –Human endothelial cells –Human monocytes –Human keratinocytes –Human fibroblasts –Human proximal tubule kidney cells –Human small airway epithelial cells Biotransformation competent cells –Primary rat hepatocytes –Primary human hepatocytes Assay formats –Cytotoxicity –Reporter gene –Gene expression –Biomarker production –High-content imaging for cellular phenotype Protein families –GPCR –NR –Kinase –Phosphatase –Protease –Other enzyme –Ion channel –Transporter Assay formats –Radioligand binding –Enzyme activity –Co-activator recruitment Cellular Assays Biochemical Assays Assays (n = 467) Chemicals (n = 320) http://www.epa.gov/ncct/toxcast/ Judson et al EHP (2010) 11

13. Signature Derivation for Rat Liver Carcinogens 12

14. Office of Research and Development National Center for Environmental Assessment 13 Virtual Tissues, Organs and Systems: Linking Exposure, Dosimetry and Response Liver Injury Tissue Morphology changes Cell Fate Transitions death /division Molecular Network Structure & Dynamics Molecular interactions & fluxes Intra/inter- cellular signaling/ fluxes Cell spatial interactions Lobular / vascular damage

15. Office of Research and Development National Center for Environmental Assessment 14 Challenges and Opportunities Extrapolation from in vitro to in vivo Recapitulation and modeling of complex cell-cell and tissue interactions. Development of virtual models to describe systems biology Recapitulation of complex behaviors

16. Office of Research and Development National Center for Environmental Assessment 15 A pilot implementation of a new approach for risk based decision- making, including characterization of risk management needs, policy relevant questions and implications for NexGen risk assessments; An operational scale knowledge mining, creation and management system to support risk assessment work and interface with gene environment data bases. Develop approaches using HT/HC data for toxicity pathways to predict/estimate points of departure for assessment purposes. Prototype examples of increasingly complex assessments responsive to the risk context and refined through discussions with scientists, risk managers, and stakeholders. This strategy focuses on development of:

17. Screening/Ranking Tier 1 10,000s of chemicals Limited decision-making Regulatory decision-making Increasing Weight of Evidence NexGen Types of Data High Throughput Molecular Mechanisms of Action In vitro only bioassay batteries (~73-500 assays) Network/disease pattern recognition Metabolism or surrogates QSAR Anchored to in vivo data Bioinformatic data integration +High Content/Med Throughput Adds Tissue/Organism Level Integration Short-term in vivo exposures with in vitro assays Mammalian species Alternative species Primary tissue culture In silico virtual tissues In vivo or anchored to in vivo data Bioinformatic data & knowledge integration +High Content, Med/Low Throughput Adds Most Realistic Scenarios Molecular epidemiology & clinical Studies Molecular biology + traditional animal bioassay Environmental exposures Upstream & phenotypic outcomes Mechanism of action for multiple stressors Knowledge integration Tier 2 1000s of chemicals Tier 3 100s of chemicals

18. Goals 1.Rank/ group chemicals 2.Assessment of high priority chemicals Are there existing assessments (hazard id & dose response), based on in vivo data, that can be utilized? Are there in vivo data to inform qualitative hazard? Decision Framework for Incorporating High Throughput Data YES NO Are there non-in vivo data to inform qualitative hazard? Overall WOE for hazard NO YES Assemble WOE by: Proximity to in vivo condition: tissue explants > cells in culture > cell-free assays > in silico Traditional WOE criteria e.g. multiple studies/laboratories, multiple dose-response. NO Use (Q)SAR and read- across to predict estimates of risk based on surrogate(s) and/or Relative potencies and/or dose-response YES NO Identify the chemicals of interest, exposure sources and pathways. What tissues/cell types/toxicity pathways are affected by the chemical in question? Conduct literature search to determine if new data will significantly alter existing assessment; update if needed. Use existing assessments to anchor in vitro /in silico analyses, if appropriate. ToxCast/ToxPi and reverse dosimetry Predictive Phenotyping Traditional DR modeling (w optional test data) Is data sufficient to determine relative potencies or dose-response? Assess dose-response: Conduct high throughput testing with a battery of assays Conduct alternative species &/or targeted in vivo testing (optional) Conduct high throughput testing with a battery of assays, alternative species ToxCast/ToxPi and reverse dosimetry Predictive Phenotyping Traditional DR modeling (w optional test data) how

19. Incorporating CSS/Next Generation of Risk Assessment (3-5 yrs) Three Assessment Tiers — Informed by Molecular & System Biology - Responsive to Risk Context Flagged for Additional Analysis Tier 1 Assessments Screening & prioritization Unknown hazard but exposures Thousands of chemicals High-throughput & QSAR-driven Minimize false negatives Decision-making Testing NTP, REACH, TSCA, etc. Input to Decision-making Testing, Research, Assessment Loop Tier 2 Assessments Narrow scope decision- making Limited hazard &/or exposures Many chemicals (hundreds of chemicals) High-and medium throughput assays & some systems level integration Science-based defaults & upper confidence limit risk estimates Tier 3 Assessments Broad scope, major regulatory decision- making Highest national hazard & exposures Few chemicals (dozens) All feasible, policy- relevant emerging & traditional data Best estimates of risk & uncertainty analyses 15 Research by NCCT, ORD labs, & partners Predictiv e Systems Models PPRTV’s & IRIS Superfund tech center & PPRTV’s IRIS, ISA’s & Multi- Pollutant Assessments

20. Office of Research and Development National Center for Environmental Assessment 19 Toxicity Pathways in Prioritization Toxicity Pathways in Risk Assessment Institutional Transition The Path to 21 st Century Toxicology

21. The Future of Risk Assessment Summary 24 The landscape of risk assessment is changing to an extent that significant modernization of risk assessment is necessary. These changes are driven largely by advances in understanding the gene environment; the important input and advice from expert science panels; and volumes of new test data from Europe. These events prompt us to look anew at risk assessment and develop this strategy to thoughtfully position environmental health scientists and assessors for the future and contribute to meaningful change within the larger risk assessment/risk management community. The goal of this strategy is to map a course forward, focusing on creating 1st approximation NexGen risk assessments, learning from these efforts and, then, refining the next versions based on this new knowledge. It may take a decade before risk assessment can rely primarily on new advances in science It is necessary, however, to begin now to address needed changes.

22. Figure by Jane Ades, Courtesy National Human Genome Research Institute Thank you